SQL Query - Actual performance duration for every execution

  • Hi,

    I have a SQL query with couple of nested queries (query at the bottom for the reference), now the very first time I execute the query on a SQL Server instance, it takes around 50-60 seconds to execute, but there on for every consecutive execution, it barely takes 1 second and I am baffled, because I want it to take its original time for me to do anything about reduction of its original execution time. I want to try various tuning and query modification aspects.

    I have tried clearing all caches using the commands listed underneath, but they does not help in getting the original execution time.

    We are using "SQL Server 2012 Enterprise Edition"

    use <Database_Name>

    GO

    CHECKPOINT;

    GO

    DBCC FLUSHPROCINDB(<Database_ID>)

    DBCC FREEPROCCACHE

    DBCC DROPCLEANBUFFERS

    DBCC FREESYSTEMCACHE ('ALL')

    DBCC FREESESSIONCACHE

    GO

    Normally anyone would be happy that query is performing great after first execution but my aim is different, I want it to take actual time for every consecutive execution

    Please help me to resolve the issue mentioned above, any pointer would be great, please let me know if you need more details

    thanks,

    Mrinal

    Query Text for Reference

    =======================

    SELECT

    OprtngOrigHandlingNodeID,

    OprtngDestHandlingNodeID,

    ShpmtOrigHandlingNodeID,

    ShpmtDestHandlingNodeID,

    CAST(COUNT(*) as real)/ (

    SELECT ABS(DATEDIFF(d,'2013-02-01','2013-02-28'))+1 as DaysInRange

    ) AS cars

    FROM (

    select distinct S.*

    from RBIS_UI.Trip_Plan TP,

    RBIS_UI.combined_shipments S,

    (

    SELECT *

    FROM (

    SELECT T.TrainRunID,

    T.Rundate,

    T.TrainRunCode,

    T.Distance,

    T.Speed,

    T.Cars,

    T.InterModal,

    T.Length,

    T.Weight,

    T.WorkStops,

    T.NonDesignStops,

    T.HP,

    T.HPTT,

    T.ActLocoCount,

    T.DeactLocoCount,

    T.TotalLocoCount,

    T.DepDesignTimeDiff,

    T.ArrDesignTimeDiff,

    T.TransitTime,

    T.DesignTransitTime,

    T.LocalArrtime AS Arrtime,

    T.LocalDeptime AS Deptime,

    T.LocalDepSchedTime as DepSchedTime,

    T.LocalArrSchedTime as ArrSchedTime,

    T.DepDelay,

    T.ArrDelay,

    T.titanID,

    T.tyesID,

    T.CorCapacity,

    COALESCE (dt_TrainName, trainname) as trainname,

    COALESCE (dt_OpsTypeId, OpsTypeID) as OpsTypeID,

    COALESCE (dt_BizType, BizType) as BizType,

    COALESCE (dt_ActualOrigHandlingNodeID, ActualOrigHandlingNodeID) as ActualOrigHandlingNodeID,

    COALESCE (dt_ActualDestHandlingNodeID, ActualDestHandlingNodeID) as ActualDestHandlingNodeID,

    ActualOrigHandlingNodeID as Actual_ActualOrigHandlingNodeID,

    ActualDestHandlingNodeID as Actual_ActualDestHandlingNodeID,

    DesignOrigHandlingNodeID,

    DesignDestHandlingNodeID,

    OpsTypeID as Actual_OpsTypeID,

    BizType as Actual_BizType,

    TrainName as Actual_TrainName,

    T.DesigntrainID,

    DETOURFLAG,

    Via,

    CarsLoaded,

    IntermodalLoaded,

    CarsEmpty,

    IntermodalEmpty,

    Addition -- ,

    -- DepDesignTimeDiff,

    --ArrDesignTimeDiff

    FROM (

    SELECT *

    FROM (

    SELECT * FROM RBIS_UI.train_runs

    WHERE RunDate BETWEEN '2013-02-01' AND

    '2013-02-28'

    ) FT

    WHERE OpsTypeID IN ( 1,2,3,5,6,9 ) ) T

    ) TR

    WHERE ((TR.TrainName = '470')) AND ((TR.ActualOrigHandlingNodeID = 1190)) AND ((TR.ActualDestHandlingNodeID = 504)) AND ((TR.DesignTrainID = 1901)) AND ((TR.OpsTypeID = 3)) AND ((TR.BizType = 'SECO'))) TR

    where TR.TrainRunID = TP.TrainRunID

    and TP.CSN = S.CycleSerialID

    and TP.Shipment_ReleaseDate = S.ReleaseDate

    AND TP.EqpmtType <> 2 ) FS

    GROUP BY GROUPING SETS (

    (OprtngOrigHandlingNodeID),

    (OprtngDestHandlingNodeID),

    (ShpmtOrigHandlingNodeID),

    (ShpmtDestHandlingNodeID)

    )

  • mrinal.technology (4/16/2013)


    Hi,

    I have a SQL query with couple of nested queries (query at the bottom for the reference), now the very first time I execute the query on a SQL Server instance, it takes around 50-60 seconds to execute, but there on for every consecutive execution, it barely takes 1 second and I am baffled, because I want it to take its original time for me to do anything about reduction of its original execution time. I want to try various tuning and query modification aspects.

    I have tried clearing all caches using the commands listed underneath, but they does not help in getting the original execution time.

    We are using "SQL Server 2012 Enterprise Edition"

    ...

    Can you please post the actual execution plan (as attachments) for a slow run and a fast run? If you're not sure how to do this, there are instructions in the first link in my sig.

    Please detail how you will get a new "very first time" now that it's been run.

    The query is unnecessarily convoluted and has potentially serious design flaws. Unless this is a deliberate attempt to generate a poor plan, I'd consider rewriting it as a higher priority than performance concerns.

    “Write the query the simplest way. If through testing it becomes clear that the performance is inadequate, consider alternative query forms.” - Gail Shaw

    For fast, accurate and documented assistance in answering your questions, please read this article.
    Understanding and using APPLY, (I) and (II) Paul White
    Hidden RBAR: Triangular Joins / The "Numbers" or "Tally" Table: What it is and how it replaces a loop Jeff Moden

  • Challenge here is that Query is not performing bad, in fact after first execution it takes less than 1 second, which is good, but I want it to revert back to original duration and then I can plan re write, if various other optimization steps fail. I do not want it rely on various caches to start with. If after first execution I cannot get the original time, then there's little I can do to to optimize

    Currently, I am attaching the execution plan after first execution where it takes less than 1 second, as per a colleague, strange part is that there's no difference in the execution plan in the either case

  • Hi,

    Issue resolved, our database has the auto create and update statistics on, so first time it was slow due to statistics creation and thereon since the optimum plan was generated to it use to be very fast.

    I am planning to manually update stats to avoid the first time slow execution. Thanks for the suggestion

    thanks,

    Mrinal

Viewing 4 posts - 1 through 3 (of 3 total)

You must be logged in to reply to this topic. Login to reply